Databricks Machine Learning Professional Question 88
Select 3You are a Machine Learning Engineer managing the lifecycle of models in a Databricks workspace. Your team uses the MLflow Model Registry to track and manage models. Recently, a model in the 'Staging' stage has been marked as outdated, and your team has decided to archive it and clean up older versions that are no longer needed to save storage space. Which of the following actions should you take to efficiently transition, archive, or delete specific model versions?
- A
Transition the outdated model version's stage to 'Archived' using the MLflow Model Registry API.
- B
Delete the model version using the MLflow Model Registry API if it is no longer needed.
- C
Manually delete the model version files from the underlying storage without using MLflow APIs.
- D
Use a lifecycle management job to automate the archival or deletion of models based on predefined criteria.
- E
Directly edit the model's metadata in the Databricks workspace UI to mark it as archived.
Show answer and explanation
Correct answers: A, B, D
Explanation
Managing the lifecycle of models in the MLflow Model Registry requires using supported APIs and automation tools to ensure consistency and efficiency. Archiving outdated models by transitioning their stage and deleting unnecessary versions using appropriate APIs are best practices. Automating these processes with lifecycle management jobs can further streamline the workflow, while manual interventions or unsupported methods can lead to errors or inconsistencies.
- A. Correct.
Correct. Transitioning the model's stage to 'Archived' is a recommended approach for marking outdated models while retaining the version for reference.
- B. Correct.
Correct. Deleting a model version using the MLflow Model Registry API is the appropriate method for removing models that are no longer needed.
- C. Incorrect.
Incorrect. Manually deleting files from storage without using MLflow APIs can lead to inconsistencies in the Model Registry and is not a recommended practice.
- D. Correct.
Correct. Automating lifecycle management with a job ensures consistent archival or deletion of models based on rules, reducing manual effort.
- E. Incorrect.
Incorrect. Editing metadata directly in the UI is not a supported or recommended method for managing the lifecycle of models in MLflow.